###
DOI:
本文二维码信息
基于差分隐私和集成学习的可再生能源出力预测方法
李洋1, 王臻懿1, 谈竹奎2
(1.澳门大学智慧城市物联网国家重点实验室;2.贵州电网公司电力科学研究院)
Renewable generation prediction method based on differential privacy and ensemble learning
Li Yang1, Wang Zhenyi1, Tang Zhukui2
(1.State Key Laboratory of Internet of Things for Smart City,University of Macau;2.Guizhou ElectricSPowerSResearchSInstitute)
摘要
相似文献
本文已被:浏览 2次   下载 0
投稿时间:2024-10-19    修订日期:2024-10-28
中文摘要: 可再生能源对于保障能源供应、改善气候环境以及调节国家能源结构都具有至关重要的作用,并且准确的可再生能源出力预测有利于维持系统平衡和降低运行成本。然而,由于可再生能源的出力高度依赖气候环境,具有复杂的时空相关性,因此传统方法难以进行准确的预测。尽管机器学习模型逐渐被广泛应用于可再生能源出力预测中,但是通常会面临模型训练数据不足的问题,从而阻碍准确预测的实现。为了解决上述问题,本文提出了一种基于差分隐私和集成学习的可再生能源预测方法来实现准确预测。具体来说,我们首先通过提出基于差分隐私的数据保护方案来实现数据的聚合,从而解决数据不足的问题。此外,我们还提出了一种基于集成学习的可再生能源出力预测框架,利用多个基本模型来共同完成预测,从而提高预测的准确性和鲁棒性。案例研究全面证明了所提出方法的有效性和优越性。
Abstract:Renewable energy plays a vital role in ensuring energy supply, improving climate and regulating national energy structure. Forecasting renewable generation accurately is conducive to maintaining system balance and reducing operating costs. However, since the output of renewable generation is highly dependent on the climate and has complex spatiotemporal correlations, it is difficult to make accurate predictions using traditional methods. Although machine learning models are gradually being widely used in renewable energy output prediction, they usually face the problem of insufficient model training data. This type of problem hinders the realization of accurate predictions. In order to solve the above problems, this paper proposes a renewable generation prediction method based on differential privacy and ensemble learning to achieve accurate predictions. Specifically, we first propose a data protection scheme based on differential privacy to achieve data aggregation, thereby solving the problem of insufficient data. In addition, we also propose a renewable energy output forecasting framework based on ensemble learning, which uses multiple basic models to jointly complete the prediction, thereby improving the accuracy and robustness of the prediction. Case studies comprehensively demonstrate the effectiveness and superiority of the proposed method.
文章编号:20241019001     中图分类号:    文献标志码:
基金项目:
引用文本: